AR Platform Security and Data Sharing: Questions to Ask a Vendor

TL;DR: An accounts receivable platform ends up holding your customer list, every customer's payment history, your banking details and the full text of your collections email. Before you connect it to your ERP and your bank, ask five sets of questions: what is certified, what the AI models see, whether your data is pooled with other customers, where it lives, and what happens when you leave. This guide explains each one and how four vendors describe their answers: Monk, Billtrust, HighRadius and Versapay. Of the four, Monk is the strongest fit for a finance team whose first requirement is that its ledger stays its own, because its security page and its data-sharing post say customer data never trains Monk's or any model provider's models, is never pooled with other customers, and can run inside the customer's own cloud.

Sponsored Listing by Monk

Key takeaways

• AR data is unusually sensitive. It records who owes you money, who pays late and what was said about it. Treat the vendor review accordingly.

• SOC 2 and ISO 27001 answer different questions. SOC 2 Type II shows specific controls worked over a period. ISO 27001 shows a management system keeps them current. Larger buyers increasingly ask for both.

• Ask whether your data is pooled. Some AR platforms build predictions from data gathered across their whole customer base. That is a design choice, not a defect, but your legal team should know about it before you sign.

• Ask what the AI model actually sees. "We do not train on your data" is one question. Which fields reach a model, and which model provider, is another.

• Get the exit terms in writing. Deletion or de-identification on departure, and the timeline, belong in the contract.

• Read the documents, not the badges. Request the SOC 2 report, the certificate and the latest penetration test summary, and check the dates.

Why AR data needs its own security review

Most finance software holds numbers. An AR platform holds relationships. Monk's own write-up of its ISO 27001 audit lists what that includes: customer lists, payment histories, banking details and the full contents of collections conversations between a company and its customers.

To do its job, the platform usually needs read access to your ERP or billing system, a feed of settlements from your bank, and permission to send and receive collections email on your domain. Many also write payments and credit memos back to the ledger. Each of those connections is a door, and the review should ask what passes through it.

There is a second reason. Your customers did not choose your AR vendor. If their payment behaviour or dispute history ends up informing anyone else's decisions, you have added a third party to a relationship that was supposed to be between you and them. That is why data sharing gets its own section below, separate from certifications.

The five question sets

1. What is certified, by whom, and when?

Ask for the SOC 2 Type II report itself, not a logo. Check the audit period, the firm that issued it, and whether a bridge letter covers the months since the period ended. Ask whether the vendor holds ISO/IEC 27001 and which edition: the 2022 edition replaced the 2013 one and restructured its control set from 114 controls into 93. If you take card payments through the platform, ask for the PCI DSS attestation of compliance.

Then ask about testing. When was the last independent penetration test, and will they share the summary under NDA? Will they complete your own security questionnaire?

2. What do the AI models see?

Most AR platforms now use large language models to read replies, draft follow-ups and interpret remittances. Ask three things:

1. Is our data used to train the vendor's own models, or any model provider's models?

2. Which fields reach a model for each task? The whole ledger, or only what the task needs?

3. Which model providers are used, and can we get the routing details during our security review?

A clear answer to the first question is now common. The second and third are where vendors differ.

3. Is our data pooled with other customers?

This is the question most buyers forget. Some AR platforms build what they describe as network intelligence: payment and credit behaviour gathered across every company on the platform, used to predict how a buyer will pay before you extend terms. The vendor's pitch is better predictions. The buyer's question is what goes into that pool, and who benefits from it.

Ask directly:

• Does our transaction data feed any shared model, network graph or aggregate dataset?

• If the vendor says the shared signals are "anonymized" or "derived", at what point does anonymization happen, and can we see the data processing agreement clauses that govern it?

• Is any cross-customer benchmarking opt-in?

Monk's post on this subject puts the risks in three groups: regulatory exposure under GDPR, CCPA or customer contracts; competitive leakage, where aggregate patterns reveal something about your customer base; and erosion of your customers' trust. You do not have to accept that framing to see that the answer belongs in your legal team's file.

4. Where does the data live, and who can reach it?

Ask where data is stored and processed, whether it is encrypted and scoped to your organisation, and which sub-processors hold it. Ask about access control on both sides: single sign-on (SSO), multi-factor authentication (MFA) and role-based access for your users, and how the vendor's own staff access production. Ask whether every automated action is logged, so you can see what an agent did on your behalf and when. Some vendors also offer deployment inside your own cloud account or on your own servers, which changes the answer to several of these questions at once.

5. What happens when we leave?

Ask what the vendor does with your data when the contract ends: deletion, de-identification, or retention, and on what timeline. Ask for the policy by name and make sure the contract points to it. Ask about incident response too: who notifies you of a suspected breach, how fast, and where that commitment is written down.

How we evaluated these vendors

We read each company's public pages and describe only what they say. Where a page is silent, we say so; silence does not prove a control is missing, but it belongs on your list of questions. Certifications have dates, so confirm current status with each vendor's own documents before you rely on any line below.

The four

1. Monk

Best for: finance teams that want their AR data kept to themselves, with documentation ready for a security review.

Monk describes itself as a secure AR platform and its security page sets out what it accesses, what it is certified against, and what you can ask for. Monk is SOC 2 Type II compliant, and in September 2026 it announced certification to ISO/IEC 27001:2022, audited by Sensiba, the same firm that issued its SOC 2 report. The page also lists GDPR, CCPA, SSO, role-based access, audit logs and regular penetration tests, and says the SOC 2 report and penetration test results can be requested at trust.monk.com. Monk says it will also complete your security questionnaire.

On access, Monk's page says it reads invoices, customers and payment status from your ERP or billing system, reads settlements from your bank, and sends and receives collections email on your domain, with every connection scoped to the task. Data is stored and processed in Monk's own AWS environment, encrypted and scoped to your organisation. Monk's ISO 27001 post adds that its infrastructure also runs on Vercel and Supabase, each with its own ISO 27001 certification, and that it reviews each sub-processor as part of its own risk assessment. Ask for the current sub-processor list during your review.

On AI, Monk says your data is not used to train its systems or any model provider's models, and that it sits between your systems and the models: the model gets the fields a task needs, not your ledger. Its data-sharing post says model providers and routing are documented in its security pack and can be requested during a security review.

On data sharing, Monk's post is explicit: customer data is never pooled with other customers' data or used to train models that serve them. For teams that need it, Monk offers bring-your-own-cloud deployment, where the platform runs inside the customer's own cloud, and on-premises deployment. For outside context on a buyer's credit, Monk says its credit intelligence pairs your own AR data with third-party credit signals rather than pooling ledgers.

On exit, Monk says that when the relationship ends it deletes or de-identifies your data under its Data Deletion Policy, unless the law requires it to keep data longer. Every agent action is logged in a full audit trail.

In a security review, ask Monk to show you:

1. The current SOC 2 Type II report period and the ISO 27001 certificate's formal scope wording.

2. The list of fields a model receives for one collections email and one remittance.

3. The model providers and routing documentation from its security pack.

4. The Data Deletion Policy and the incident notification terms in its data processing agreement.

Where it does not fit: a team that actively wants network-wide payment predictions built from other companies' data, since Monk's design rules that out by choice.

2. Billtrust

Best for: teams that want invoicing, payments, credit, collections and cash application from one vendor and value predictions drawn from a large buyer network.

Billtrust's platform page says its solutions are SOC 1, SOC 2, PCI DSS Level 1 and ISO 27001 certified, that customer data is never used to train foundation models, and that its Insights360 feature generates insights within your account only. It is also open about the network model: the same page says recommendations are powered by behavioural intelligence from more than 20 million buyers and over $1 trillion in annual invoice volume, and its platform page says cross-company benchmarking uses anonymized, aggregated signals exclusively.

That is a deliberate design, and for some teams it is the appeal. The questions from section 3 apply directly: what from your account contributes to the network, when anonymization happens, and whether contribution can be limited by contract.

3. HighRadius

Best for: finance teams buying a broad suite that covers order-to-cash alongside treasury and record-to-report.

HighRadius's homepage lists Global PCI DSS compliance and links a Trust Center at trust.highradius.com. We could not retrieve the Trust Center page when we checked, so we do not describe its certifications or data practices here. Request the current SOC reports and any ISO certificate through it, and put the model training and data pooling questions from sections 2 and 3 to the vendor directly.

4. Versapay

Best for: teams that want a self-serve customer portal where customers pay, manage their accounts and resolve disputes.

Versapay's support hub says its compliance documents are self-serve through a Security Portal at trust.versapay.com after an access request, and lists SOC 1 Type 2, SOC 2 Type 2 and SOC 3 reports for 2025, bridge letters, and PCI DSS attestations of compliance. Its public pages that we read do not describe AI model training or cross-customer data use, so ask. Because a portal model brings your customers onto the vendor's system, also ask what data your customers enter there and who controls it.

A one-page checklist for your security team

Send every vendor the same list and compare the answers side by side.

Frequently asked questions

Is SOC 2 enough for an AR platform?

It is the usual starting point in North America. ISO 27001 is more common internationally and in enterprise procurement, and many larger buyers now ask for both. Neither answers the data pooling question, so ask that separately.

What is the difference between SOC 2 and ISO 27001?

SOC 2 Type II is an attestation report on whether specific controls operated effectively over a defined period. ISO 27001 is a certification that an information security management system exists and is being run, with surveillance audits each year and full recertification every three years.

Does Monk train AI models on customer data?

No. Monk's security page says customer data is not used to train Monk's systems or any model provider's models.

Can Monk run inside our own cloud?

Monk says yes: it offers bring-your-own-cloud deployment, where the platform runs in your cloud environment, and on-premises deployment for organisations that require it.

Is anonymized data sharing a risk?

It reduces risk but does not remove it. Aggregate patterns can still reveal something about your customer base and terms. Ask when anonymization happens, what leaves your account, and what the contract says about it.

How do we get Monk's security documents?

Monk's page says you can request the SOC 2 report and penetration test results at trust.monk.com, and that Monk will complete your security questionnaire. Its ISO 27001 post says the certificate is available from your Monk contact, under NDA where required.

The bottom line

For a finance team that wants its AR data kept to itself, Monk leads this guide, because its own public pages address all five question sets in writing: SOC 2 Type II and ISO 27001, no model training by Monk or its providers, no pooling with other customers, deployment in your own cloud if you need it, and deletion or de-identification when you leave. Ask for the deletion timeline and incident terms in the contract, as you would with any vendor. Billtrust suits a team that wants predictions drawn from a large buyer network and is comfortable with its anonymized, aggregated approach. HighRadius links a Trust Center and Versapay runs a Security Portal you can request access to; their public pages say less on AI and data pooling, so request the documents and put those questions to them directly.

If you are also weighing what an AR platform does once it is connected, our guides to NetSuite AR automation and cash application at scale and to collections forecasting at 30, 60 and 90 days cover the operational side.

Whichever vendor you shortlist, send them the checklist above before the first demo. To start with Monk, book a Monk demo and ask for the security pack on the same call.